学步园

17

FSIS

Can CTS be used to feed multiple instances of ESP?

Yes, CTS can feed multiple instances of ESP e.g. based on rules. BUT, in contrast to feeding a single ESP installation (where state is tracked, call backs provided, etc.), when setting up multiple parallel ESP installations, CTS will OOTB not keep nor monitor state of the feeding process, ensure recovery and synchonization, and provide all functionality that the ESP feeding proxy is traditionally offering.

32

FSIS

Can I have Crawler -> CTS -> ESP

Yes, CTS act as a CD and can be put in the feeding chain between the crawler and ESP Content Distributor.

22

FSIS

Can I use my existing modules from Unity in IMS?

No. Unity was written in Python. The primary extension point for IMS (Run Code operator) is C#.

21

FSIS

Does the IMS UI Toolkit contain webparts to integrate into SharePoint?

No. For V1, no webparts will be included. This might change for future releases.

1

FSIS

How do A/B testing capabilities of IMS compare to SUMO?

First of all, the IMS A/B testing capabilities in FSIS will not be on par with those of SUMO. IMS will only allow you to control exposure of a site configuration based on user distribution (ie. 5% of users hit site config A while 95% of users hit config B). SUMO, on the other hand, will offer far greater control over exposure: User distribution (like IMS), Browser targeting, OS targeting, Based on referral URL, Mobile browser vs. desktop browser, etc.

In addition to more granular control of exposure, SUMO also offers measurement of the performance of a particular configuration (click-throughs, search-result quality, etc.). SUMO also offers a psudo-UI framework by templatizing UI sections and allows these sections to be controlled through SUMO (I want “ads” on top, then “video results”, then “Image results”, etc.).

In some ways IMS and SUMO overlap. IMS is based on a very different framework, and might have more potential than SUMO to become a rich and user-friendly UX orchestration suite. But, IMS 1.0 is not a SUMO replacement (neither was it intended to be). IMS is not a OOTB A/B testing solution either. IMS offers excellent capabilities for controlling how search-driven applications are populated with content (+ the UI toolkit, of course)—and that’s the area one should elevate when promoting IMS.

5

FSIS

How do you backup an FSIS installation?

Backup of the ESP component is (should be) covered in the ESP documentation. With respect to IMS and CTS, custom backup of the flows is performed by simple file backup (or backup of a source-control tool if one uses that). The engines do not support recovery, so there is no point in trying to backup state of those. The engine bits and studio should be re-installed when recovering.

19

FSIS

How does CTS handle authentication for data sources?

The native readers (e.g. DB reader) are configured with access parameters and credential. These credentials are stored as flow property (encrypted) and are deployed together with the flow.

2

FSIS

Is FSIS interoperable with older versions of ESP?

OOTB integration of CTS and IMS is only supported for ESP 5.3 SP3.

For IMS-federation to older ESP-versions you should use the RunCode operator. It is also fully possible to _try_ to use the OOTB ESPLookup operator towards any ESP-version compatible with the 5.3 SDK, but you do it at your own risk--this will not be supported by R&D.

4

FSIS

Processing in CTS vs. ESP pipeline: When is processing in CTS preferable over the ESP pipeline and vice versa?

Scenarios where CTS processing is preferable:
Content Fusion: Use CTS to create the content unit (document) that you want to index. For example: Enrich unstructured content with structured content before indexing.
RDBMS: CTS is pretty good at extracting and transforming structured content, so whenever you have an RDBMS source you can use CTS to preprocess the data before feeding ESP. In addition, when dealing with non-ESP targets, CTS is an obvious alternative.
Other scenarios for CTS:
- Parsing XML: CTS has a pretty good XML parser. Use this as an alternative to coding a custom connector (i.e. the source system exports to XML, and CTS parses it).
- Custom extraction based on reg-ex or dictionaries.
- Set-oriented operations: When you need to perform processing across record/document sets
- Noise removal when feeding HTML. CTS contains RCD (Relevant Content Detection) for HTML which helps identify the main text of the record so that you can remove typically unwanted elements like ads and navigation.
- Content quality: The fuzzy matcher operator provides high-speed records linkage functionality that can be used for content quality enhancements.

Scenarios where pipeline processing is preferable:
The pipeline has a more complete feature set for linguistics and OOTB extractors. Also, the pipeline offers complete access to the matcher framework while CTS only caters for RexEx and disctionary-based matching (verbatim + levenshtein). Following from this, use the pipeline for:
- Access to the matcher framework
- OOTB extractors
- Linguistic processing (Vectorization, NP extraction, Lemmatization, etc.)
- Link-cardinality analysis (web analyzer)
- Classification (RBC + FSA)
- Robust feeding of multiple deployments (feeding proxy)

34

FSIS

Query reporting in FSIS: How are IMS query logfiles consolidated (presumably to a FAST admin node)? How is query reporting provided to end-user customers in FSIS? (eg how is Flare / Clarity replaced?  SSRS?)

In short, the query reporting will not change with the introduction of IMS. There is no OOTB query logging in IMS; you can easily design custom logging through using a DBWriter (for example) as a part of your IMS flow.
 
In addition, the ESP query reporting framework will still log everything—the same as with ESP today. Through the query reporting framework you can export the contents of the log files to an RDBMS, and put any reporting solution on top of that (for example SSRS)—the same as with ESP today. Clarity and Flare can also be used going forward.

9

FSIS

Sizing metrics: How many docs/sec can you process on one CTS node/constellation given a certain document type and processing complexity?

Guidelines on sizing metrics will be made available as soon as R&D has completed benchmarking with optimized code.

8

FSIS

Sizing metrics: How many QPS can you get out of one IMS node/constellation given a certain query processing complexity?

Guidelines on sizing metrics will be made available as soon as R&D has completed benchmarking with optimized code.

18

FSIS

The FAST Search Designer looks like a very separate component. What are the requirements to use it?

The Designer requires that a FSIS back-end (IMS/CTS) install is present on the same machine. A typical development deployment consists of a single host install of CTS/IMS, on top of which the Designer is installed. This does not mean the Designer can only talk to locally installed CTS/IMS engines. It can be pointed to engines installed remotely as well.

15

FSIS

What "pre-built" webcontrols will be shipped in the IMS UI Toolkit?

We offer two types of UI components - pure Ajax/JavaScript and .NET Controls. The plan is to provide the following:

Ajax/JavaScript: Chart Navigator, Text Navigators, DidYouMean, Document Fields, Document View, Hit List, Navigation Bar, Progress (wait icon while searching/rendering), Refinements, Results Panel (a behavior that hides elements when searches are inactive), and Tag Cloud. There are also some non-visible components - Search, SearchBuilder are the two primary ones.

.NET Controls: Chart Navigator, Text Navigators, DidYouMean, Document Fields, Document View, Hit List, Navigation Bar, Progress, Refinements, Results Panel (as an extender), and Tag Cloud. Also include Search and Search Builder.

Planned but at-risk items: Taxonomy Navigator, fully fleshed out Chart Navigator

12

FSIS

What distance algorithms for fuzzy matching are available in CTS?

CTS has two operators that cater to fuzzy matching: 1) The fuzzy joiner and 2) the dictionary lookup operator.

The fuzzy joiner is designed for records linkage jobs. I.e. when you want to investigate if two records are “related” (data cleansing). The fuzzy joiner only provides n-gram based algorithms in v 1.0. Other algorithms are under consideration for future releases.

The dictionary lookup operator is a “matcher” that lets you perform exact or fuzzy matches against a dictionary. The fuzzy algorithm used in the dictionary lookup is Levenshtein.

6

FSIS

What ESP connectors will be officially supported in FSIS?

​

3

FSIS

What is the total list of operating systems which are officially supported in ESP 5.3 SP3?

​

10

FSIS

What kind of query classification capabilities are available in IMS?

​

13

FSIS

What tools for dictionary management are available in FSIS?

LingStudio can be downloaded for use with the ESP dictionaries.

CTS/IMS offers a dictionary compiler operator which allows one to use the framework to set up flows that produce dictionaries. The entire notion of dictionaries change with CTS and IMS. What matters more, is the source data: With access to the source data, CTS and IMS are very flexible wrt dictionaries.

Important: Think of CTS/IMS as a different ecosystem from ESP wrt. dictionaries. We do not guarantee nor support interoperability between the automaton formats between these components. Use ESP-tools to manage ESP dictionaries. Use CTS/IMS tools for CTS/IMS dictionaries.

20

FSIS

Will connectivity from the BDC accessible for CTS?

No. There is no BDC bridge in CTS for V1.0 because the tight coupling of BDC and SharePoint.

33

FSIS

Will CTS support the Feeding Proxy

CTS can feed to the feeding proxy. CTS uses the ContentAPI, and can write to what looks like the CD.

7

FSIS

Will FSIS 1.0 be localized?

No, but future versions are expected to be localized.

14

FSIS

Will FSIS have a "silent install" option?

Yes, we plan to offer a silent install option for CTS and IMS.

16

FSIS

Will the FSIS beta be downloadable by customers.

The FSIS beta is "by invitation only". There will be a managed program with 6 customers, getting implementation support from FGS. In addition, there will be an unmanaged program without support for free, available to a broader audience of customers, but still by invitation only.

Similarly, there will be a managed and an unmanaged beta for selected partners.

37

FSIS

Will there be FSIS web parts for FS4SP?

Currently there are no plans for “FSIS Webparts” being a part of the IMS UI Toolkit. The UI toolkit does offer ASP.Net controls, and you could use these to build custom SP webparts that talk to IMS.

If you don’t want to use the UI Toolkit, then you have to write your own WebParts that speak http directly with IMS (+ the FSIS Admin service).

11

FSIS

Will training-based or rule-based classification be supported in CTS?

No training-based in CTS. No Rules-based classification operator either. But one can easily use CTS to model simple (well, you can actually model quick advanced stuff, but it becomes unmanageble very easily) rules-based classification schemes. BUT, CTS is is not designed to replace RBC or FSAClassifier in version 1.

Recommendation: use FSA or RBC using the pipeline. For simple rules-based models, feel free to try CTS.